AI with Kyle Daily Update 114
Today in AI: The Capability Overhang
Watch this video on our dedicated watch page
Better viewing experience with related videos and full-screen player
The skinny on what's happening in AI - straight from the previous live session:
Highlights
Weird day today because basically there’s no news! All the AI companies have been quiet over the holiday period - expect it all to kick back off next week.
Instead of just writing some fluff I used the Livestream to dig into a concept called the “capability overhang” which is going to dominate 2026.
Andrej Karpathy - one of the guys who literally founded OpenAI, former head of AI at Tesla, genuine legend - posted something this week that made me sit up.
He said he's never felt this behind as a programmer.
Let that sink in. One of the people who built this technology is struggling to keep up with it.
If Karpathy's struggling, what hope do the rest of us have?
What we’re talking about here is a capability overhang - something I’m super interested in…let’s dig in.
In a similar vein…a few days before Christmas, OpenAI put out a statement. And it wasn't about some new model or product launch.
They said that progress towards AGI in 2026 will depend as much on helping people USE AI effectively as on building more powerful models.
Read that again.
OpenAI - the company that's been racing to build the most powerful AI on the planet - is essentially saying: the tech isn't the bottleneck anymore. People not knowing how to use it properly is.
They're calling it "capability overhang." The gap between what AI can already do and what people actually use it for. And they reckon this gap is "immense."
This isn't new, by the way. Ethan Mollick - the Wharton professor who wrote Co-Intelligence and is probably the most sensible voice on AI right now - has been banging on about this for ages.
Back at the start of the year he predicted that AI capability would grow much faster than two things: people's understanding of what it can do, and organisations' ability to absorb the pace of change.
And he's been proved absolutely right. The models have become extremely good this year. GPT-5.2, Claude Opus 4.5, Gemini 3 - massive leaps.
But most people are still using AI the same way they were using ChatGPT 3.5 two years ago. Write me an email. Summarise this document. That's it.
The gap between what's possible and what people actually do with it just keeps getting wider.
💰 Why This is Your Opportunity
Here's where this gets interesting for anyone paying attention.
People's minds are stuck three years ago. They're still worried about whether AI can count the Rs in "strawberry." They're still thinking AI is for writing dodgy blog posts. AI is for funny meme images.
Meanwhile, these tools can help you set up entire businesses. They can help you structure and complete complicated processes. They can smash gigantic tasks into pieces for you.
Someone needs to close that gap. Someone needs to help people catch up with what AI can already do.
Could be you?
Companies like Deloitte got caught this year using AI stupidly - releasing reports with hallucinated footnotes to government clients. Twice. That's a human training problem, not a technology problem. And now the whole consulting industry looks a bit redundant because clients are asking: if you're just using AI anyway, why am I paying you half a million quid? VERY fair question.
I charge about £1,500 ($2000) an hour for AI workshops. And honestly? It's easy work. Because most of what I teach is stuff that feels basic to people like us who've been paying attention. But it's revolutionary to the people who haven't. And that’s the majority.
If you can use AI effectively in your niche - and you can explain it to others - that's a business right there.
2026 is going to be about closing the capability overhang. Help people catch up with what AI can already do. Get paid very, very well for doing so.
Want the full unfiltered discussion? Join me tomorrow for the daily AI news live stream where we dig into the stories and you can ask questions directly.